ServiceNow President & CPO on Why AI Kills the Companies That Don't Transform | Amit Zavery | E305

Amit Zavery 0:00 Any company who wakes up and says, "I am the end-to-end orchestrator," I think they're dead. There's no clear ROI in many cases. So you'll end up failing more often because you're doing the spare part approach versus doing a very thoughtful platform first mindset. Our AI business is going to be $1.5 billion this year. Our plan was $1 billion. All of our products have become AI-native. So if you don't transform, you will get killed. A lot of leaders will say, "Oh, yeah, we have to put all this tools together. It'll take us three months or six months or nine months." Until then, nobody does any AI or nobody does cloud or nobody does this, and that is where the companies are going to die. But going and say, "Please onboard an employee and give them access to 15 different applications based on their role, and then ship them a new laptop based on where they live, plus, uh, make sure their benefits are up to date," that is taking action. Now, the systems behind the scenes have to do that heavy lift. That's what we're solving for.

Carlos González de Villaumbrosia 0:57 Hey, this is Carlos, CEO at Product School and your host on The Product Podcast. Today's guest is Amit Zavery, President and Chief Product Officer at ServiceNow. ServiceNow is the platform enterprises run their work on, with more than 75 billion workflows flowing across it every year and around $14 billion in annual revenue growing over 20%. Even with a market cap above $100 billion, the stock is down more than 30% over the past year, while its AI business is on track for $1.5 billion this year, ahead of a $1 billion plan. Amit ran product and platform at Oracle for over two decades and was a VP and General Manager at Google Cloud before joining ServiceNow. In our conversation, we cover why the market can't yet tell the AI winners from the losers and why he believes companies that don't transform will get killed. Why the idea of one company becoming the single end-to-end orchestrator for the enterprise is a fallacy. The spare part approach that makes most enterprise AI projects fail and what the pacesetters do instead. Why access is moving away from the user interface toward agents, and what taking action, not just retrieving information, actually requires. How to hold long-term conviction on platform bets while the market judges you on short-term sentiment you can't control. Let's get into it. Welcome to The Product Podcast, Amit.

Amit Zavery 2:15 Thank you. Thanks for having me, Carlos. Very excited for the conversation.

Carlos González de Villaumbrosia 2:19 The first question I have for you is about your title, because it's quite unusual. You are President and Chief Product Officer and Chief Operating Officer at ServiceNow. So what does that really mean in practice?

Amit Zavery 2:31 Yeah, it's, uh, pretty straightforward if you think about it, right? So the chief product officer basically is responsible for defining the product strategy, building the products, so product engineering, as well as, uh, making sure that we're delivering the best core capability to our customers. And the COO role is how do you operationalize all these things? So how do you really deliver value to our customers? How do you make sure company's operating at a very good scale and, and efficiently? Uh, as well as delivering, uh, this capability in a much faster pace. So combining op- uh, strategy, product delivery to execution gives us the ability to now look at it end-to-end all the way from engineering to customers at one place. So it gets less fragmented, less, uh, uh, handoffs, and you have one person kind of overlooking it so that we can make decisions faster. So that's the goal with this, uh, role is to really move fast and do things perfectly for our customers.

Carlos González de Villaumbrosia 3:26 And when I hear you explain it, it makes total sense. And I've seen a lot of organizations, uh, unify product and technology, the chief product and technology officer, but I had never seen, uh, product and ops, uh, under the same roof.

Amit Zavery 3:41 Yeah. No, I think it's a unique thing here, and I appreciate how Bill, who is our CEO, thinks about it, and it gives us the ability to-- And we see a lot of the innovation we've done over the last 18, uh, 24 months has been that way because we've been able to now bring efficiency but also alignment across the organiz- organization, right? So have product engineering, uh, experience, and then you have operations allows us to really, uh, do a great job, I think.

Carlos González de Villaumbrosia 4:07 Can you give me a, a sense for the scale of your, your team and your business today?

Amit Zavery 4:12 So ServiceNow, uh, we are $15 billion in revenue, uh, growing at 20-plus percent. Uh, our free cash flow margin is around 35, 37%. Uh, op margin is around 32%. So it's a pretty scaled business growing very fast. One of the fastest growing, uh, software companies in the world today. Uh, and we were the fastest to $1 billion, $5 billion, $10 billion, and now $15 billion, so very proud of that fact. Uh, in terms of the size of the company, uh, we are around 29... 8, 29,000 employees. Uh, and if you're in the product and engineering, the R&D team is around 15,000 people. So most of our investment in the company is to make sure we build the best products for our customers. So half-- more than half of our investment goes there, and the rest of the things, of course, go-to-market is important for us to really work with the customers to make them successful and, uh, adopt our products, and then we have all the supporting functions around it. Uh, but our goal always has been it's a very product-led, very engineering-driven culture inside company, but customer-first mindset.

Carlos González de Villaumbrosia 5:13 So I want to address one of the big elephants in the room, right? As you mentioned, you're beating expectations, growing over 20% year over year, and your market cap is over $100 billion, yet the company's down, uh, over 30% year to date from a market cap perspective, right? So I'm curious to know why do you think the market is still not rewarding that performance and that investments that you are making on, on, on the AI front?

Amit Zavery 5:40 Yeah, I think, uh, see, there is, uh, with any trans-transformation or technology shift, people start to wonder what it means for any company which has been around in that space for some time. So this shift with AI, I think the investor community is still trying to understand what this means. Is it a tailwind or a headwind for companies like ours? And if they don't understand the change yet, they're putting every software company in the same bucket. But I think we believe we are a very different kind of a platform company than other software provider, which are, which are much more vertical stacks. So we do have a lot of tailwind. Our, our AI business is going to be $1.5 billion this year and growing at a phenomenal rate, right? Our plan was $1 billion. We already increased our plan to $1.5 billion. All of our products have become AI native. So I think it takes time for the market, uh, the investor community to understand those nuances, differences, whether we are, uh, early AI company or are we remaining legacy. So if you don't transform, you will get killed, but if you transform and build the right things and do the right things, we have a lot of domain, a lot of customer base. If we give them a lot of value from AI, they don't want to go anywhere else, and that's what we're starting to see with our growth. If you look at our growth has been very, very good and accelerating, and we're seeing a lot of new businesses emerge inside ServiceNow as well. So I think it's just timing, I would say, for investors to understand the differences between different providers. Uh, and, uh, I think it's starting for them to see there's also rotations happening with what's happening with infrastructure, what's happening with, uh, uh, the, the, the technologies out there, uh, with, uh, semiconductors and memory. So there's a lot of confusion in the market today with, uh, AI. Uh, so I think it's a matter of time from my perspective, but, uh, we feel good about our long-term strategy. We feel good, good about our traction, and I don't really worry too much about a short-term changes in our stock prices. Of course, we want to continue to be appreciated for our growth and everything, but that's, uh, for investors to decide. All we can do is control what we can, and we're doing the best we can there.

Carlos González de Villaumbrosia 7:44 Exactly, and, and I'm curious about that. So from your perspective, like, how does the market cap affect the way you are planning for the future on a, on a quarterly or maybe on an annual basis?

Amit Zavery 7:56 Yeah, I think the market cap does not directly impact... I mean, of course, we want to listen to our investors. They are a very big stakeholder, uh, for us when we make decisions. So we have a fiduciary duty to deliver returns for our investors. So we do always keep on looking at it. How can, where can we be more efficient? Where can, how can we get more growth? How do we convey and communicate our messages about what we're doing and how we differentiate it? How do we show, uh, our traction in the market? Some of the investments I'm doing ar-around new opportunities, which are going to be big roadmap opportunities for us, accelerate our TAM, right, in the security space, data space, what we're doing in the CRM space, are all driven to create more value to our customers, but also in the end for our investors. So, uh, we always think about that on a regular basis to make sure that our investors remain happy, they get great returns, but we also do the right thing for long term as well. The short term, you can have these ups and downs, but long term, the trajectory has to be great, and our trajectory has always been good, both from a financial metric, uh, innovation metric, uh, as well as, uh, return to investor from the financial metric as well, right, for

Amit Zavery 9:10 them.

Carlos González de Villaumbrosia 9:11 And so as I try to unpack that for, for product leaders specifically, there's always this tension between optimizing for maybe short-term revenue, long-term growth, right? So as, as you try to resolve some of those friction points when you, when you plan your strategy, when you expand into different products, when you try to, um, may-maybe increase the performance of your core products, like, I'm curious to know, like, if there's any specific framework or way for you to, to think about it.

Amit Zavery 9:38 Yeah. I think, see, as a product leader, you always have to first have a perspective of where the long-term opportunity is. Build things which are going to be sustainable. So my goal with any com-- uh, anything we decide is a sustainable growth. Don't do short-term things and then you have to repi-- uh, redo everything ag-every time, and our customers will be unhappy, our engineers and product leaders will be unhappy, and eventually you're not going to get the results. So I always want to balance, no doubt, investor sentiment and, uh, feedback from the c- uh, the, from the shareholders and others. But we, we also have to have conviction about what we do. You have to feel confident about where the market is going and whether, what your differentiation is going to be and how you're going to build a lot of value, uh, in our products for customers to be really excited about buying, using, and continuing to expand the usage. Uh, so short term is good to kind of pay attention to, but you cannot lose sight of long term. You have to-- The product development, even though people say that, you know, it's easy and fast to build products now because of software has become easier to build, it's not about the speed of engineering. It's always the s- the real quality comes from understanding what to build and delivering that in a much more valuable and usable way. And that's really what I focus on from a team perspective, uh, and, uh, ensure that we're getting constant feedback from our customers, our users, our partners, our ecosystem, and then have a gut feel. You have to know what's happening and where you will make the biggest, uh, investments and how you're going to monetize those things long term.

Carlos González de Villaumbrosia 11:14 Yeah. And, I mean, you've been around the block. I see you spent ton of time at Oracle, then as a VP at G- and GM at, at Google, right? So you've seen kind of some patterns around this type of internal transformation. So I'm curious to know what are some of those elements that you think stay true now in this new AI transformation, and, and what do you think it's fundamentally different from the way you've been transforming companies before?

Amit Zavery 11:39 Yeah. I think, uh, see, the, the basics are still the same, right? The best-- Building the best products and customers loving your products, you're gonna win. You have to get that right. You cannot build, uh, a sustainable business on substandard products or features which are not going to add value. So those are basic capabilities and basic things. You just feel transformation going on. Any new technology comes in, still have to do the basics right. So you have to have a very good team. Uh, people who understand how to build very scalable, secured, performant, but very nicely usable products. You have to really ensure that you have customer relationships so that you understand what they're doing, how they're doing things. You have to get input from various different things or different environments so that you can make the right decisions. So those basics remain the same. I think every transformation has introduced new, new challenges, right? When we went from, say, uh, a client server to, uh, more of a distributed architecture, web-based architecture, the technology transformation was huge. You had to rewrite software, you rebuild a lot of things. How people use your, uh, products changed quite a lot. Uh, browser-based interface is very different than what you had with the thick clients. Uh, and the data being residing in a distributed environment was very different. So it was a big technology shift. Commercially, it didn't change much. When we moved from, say, uh, a, a web-based to cloud-based app, a client-base-- a cloud-based architectures where things are hosted in a cloud, uh, they're running, uh, and you're operating those products yourself as, as an engineer. So your technology changed, the role changed, but also commercial models started changing. People were doing more subscription-based pricing versus a perpetual license pricing. So suddenly you're doing technology transformation and a commercial transformation. That you had to kind of figure out how to make that change for your customers, how do you change your sales teams, how you change the incentives, and then how do you really monitor and meter all those things, because you're also charging for, uh, u- number of users and things like that. When we are now moving to the world of AI, it's going to a much more consumption-driven architecture. Architecture is very different with AI, of course, right? You're using, uh, different large language models, using, uh, technologies for reasoning. You're bringing that into deterministic workflows. So the tech stack is changing quite a lot. But commercially also, people are now saying, "You know what? I want to do usage-based pricing. I want to do consumption. When I use it, I want to pay you." So commercially, instead of subscription, now you're going to a more consumption-driven, uh, pricing. So again, so, so there are a lot of-- every technology shift brings different, different nuances that requires the kind of people, the way you work, where you operate, where you sell, where you build, all changes around you. So basics might be the same in terms of you want to build the best products, you want to get customer satisfaction, and you want to do long-term thinking, but how you do it is changing every time, and the speed is very different now, right? What is happening with AI, the compression and time is very high. You're doing changes on a regular basis versus maybe six-month plans. You're delivering things much more in, uh, much more regularly, and you have a less time to really, uh, do, uh, long-term features. You're building things much faster now. So I think it's-- that's, I think, the change, and I think you have to be cognizant of it and deliver to those kind of thinking now.

Carlos González de Villaumbrosia 15:10 So you recently released a, a report on AI maturity in the enterprise, and I want to dig into that because I think it's a lot of companies that, or a lot of leaders will resonate with this. Uh, one of the highlights you mentioned is the AI spend is, is going up, as, as you mentioned.

Amit Zavery 15:26 Yeah.

Carlos González de Villaumbrosia 15:26 Teams buying tools or now using more, more tokens. But there, there is a constant theme there, which is, uh, a lot of those teams struggling to show real ROI. And then you mentioned that there is like a subset of companies, you call it the pacesetters.

Amit Zavery 15:41 Yes.

Carlos González de Villaumbrosia 15:41 Small percentage of companies that are somehow reorganizing for AI and not just around it.

Amit Zavery 15:47 Mm-hmm.

Carlos González de Villaumbrosia 15:47 So I would love to learn more about what you mean by that.

Amit Zavery 15:50 Yeah. I think what we've seen, and when we talk to a lot of our customers, and this report was based on a lot of interviews and analysis, uh, you find variations in terms of how customers are approaching the AI journey. A lot of them are like, "You know what? Everybody should go and do something with AI," without providing some kind of as a prescription or some kind of guidance. So the projects are all over the map for companies, right? Some of them are getting some success. Some, some projects are very failing. There's no clear ROI in many cases, but also there's no coherent way of thinking and providing expertise across the company, uh, organization. And then there's-- becomes a very scattered, scattered approach, but also very much like, uh, I would say spare part approach. You bring in technologies, pieces of technologies, and trying to cobble it together and make it work. And given the underlying technology is changing so fast, it's very hard to keep up for most of the organizations. So you'll end up failing more often because you're doing the spare part approach versus doing a very thoughtful platform first kind of mise- mindset. So the pacesetters, people who have been very successful, companies which we've seen, is that they are taking a little more holistic approach. They're providing, uh, a little more structure to the company. They're providing help and expertise as needed. And they're picking the projects which are going to be much more value add, as well as has the right kind of investment and right kind of people. Because AI is not just a technology shift, it's also a cultural shift inside the companies. And you have to approach this thing together as an organization and have everything behind it. You can't just say, "Go and do partial things." So not do spare parts, don't do pieces of technology. Buy a platform, work with people who understand this, identify the right projects, and make sure you put full focus on it to regularly introspect what is working, what is not working, and fixing those things on a regular basis. So having that mindset are the ones who are seeing the pacesetters. They've really differentiated themself by putting that energy behind this AI transformation enterprise-wide. Uh, and then putting all the energy with the right kind of to-- partners to make it all successful and fixing things which are not working very quickly so people don't just linger around. And those are the pacesetters we have seen really work out very well.

Carlos González de Villaumbrosia 18:12 You identified four main gaps for companies that are trying to go from curious to a pacesetter, right? And number one, you mentioned it's the data gap.

Amit Zavery 18:22 Yes.

Carlos González de Villaumbrosia 18:22 Apparently, like 71% of companies struggle with the data accuracy, access, and, and management. So how can companies unlock the power of their data?

Amit Zavery 18:32 Yes. See, data has been always the fragmented, hard to manage, and I think data will always remain fragmented. I think this idea that everything will come together in one place, it's, uh, not gonna happen in any enterprise. Very difficult. And because systems are built, the things are running everywhere, data is distributed. The best way to fix the data gap is to really have a very good, thoughtful understanding first what your landscape is. Where is what? Who owns it? What are the systems? So the getting that kind of understanding of that is a starting point. And second thing, which is most valuable, is really providing a very good integration layer. A data, data stack which gives you the ability to do, uh, without moving everything around, a federated sense to make decisions. Uh, so either through zero-copy adapters or through different data integrators, integration technologies, providing that layer and creating a semantic layer on top of it. Once you have a semantic layer with data distributed across multiple different domains and, uh, silos, you can have now a virtual view of everything without having to re-rationa- rationalise everything, move everything around, which is very painful, time-consuming, and never works because things get fragmented very fast again. So the semantic layer with a knowledge graph a- and then a context graph gives you a lot more insight and ability to do insight to action with-- while the data have been distributed. So data gap has to be fixed that way instead of re-engineering everything every time.

Carlos González de Villaumbrosia 20:02 So on that, on that note, we've had, uh, other product executives on the podcast from companies such as Snowflake or Fivetran, right? And everybody talks about the system of record. Some people try to be the system of record, some try-- some people try not to be the system of record. But what's clear is that data is living across different systems.

Amit Zavery 20:18 100%.

Carlos González de Villaumbrosia 20:19 And the, the, the effort that it takes to clean it up in a certain way, uh, so it's accessible for non-data teams seems to be high. So I'm curious to know from your perspective, what are some ways for teams to get the data to a point that is good enough so they can continue on their transformation?

Amit Zavery 20:36 Yeah. As I said, this idea of having, uh, a text technology which allows you to connect the systems and the data and coming from system of record, coming from data warehouses, uh, coming from file systems, all need to be available for you to make decisions. And when we-- What we have built now, we have something we call Workflow Data Fabric. The Workflow Data Fabric is basically a connectivity layer with some, uh, semantic capability to understand where your data is, how you integrate it together, but also create a semantic layer for doing insight to action, has been the stack we provide to our customers. And that has worked very well for our customers because they are able to now, uh, move very fast to get insights from all these different s-systems out there, and without having to do a lot of heavy lifting. If you ask now every data, data team to get-- go rewrite everything, move everything around, it fails. And that's why we created this Workflow Data Fabric as a technology offering with the connectivity layer, the semantic layer, and on top of that, we're building the context graph. And that has unlocked a lot of value because you are now understanding a lot of the insights from all the systems now.

Carlos González de Villaumbrosia 21:48 How do you think about the teams accessing that data? I mean, we're seeing a lot of companies going headless, right? Exposing their data via other interfaces they do not own, such as Claude, Copilot, MCP, CLI. So curious to know your perspective, because ultimately, I think in order for this to work, non-data teams should be able to use data to, to make decisions.

Amit Zavery 22:10 Yeah. I, I completely agree. So, see, the thing is, uh, access has to be now provided in multiple ways. The idea that everybody will go through a user interface is fallacy. It is not gonna work in the future. It's not-- already starting to break apart. So you have to provide, as-assuming humans are going to use your systems, agents are going to use s-systems, Copilots of the world are going to use your systems, and you need to make sure you provide access in multiple ways. Access through a user experience layer directly, but also through MCP server for that access, A2A for agent-to-agent integration, Claude Code or CoWork or Claude CoWork or, uh, Gemini Enterprise or Copilot, Microsoft Copilot. All of these things will have to have ability to take information you require to do your day job. If, if a hu-- if a human is interacting with it to do the job or agents are interacting with it to do a workflow or anything like that, that has to be available. So what w- the way we are architecting our platform today, and we have this layer called Action Fabric, is to abstract out interfaces to make it easy for every way for anyone to interface into our products, into our technology, into our data. Of course, beyond just accessing, the biggest change which we're bringing in will also be able to instruct the, uh, the systems to do action on your behalf, not just give you information. That's where the unlock happens. Getting information, say, "Hey, tell me what my latest forecast is," very simple. But going and saying, "Please onboard an employee and give them access to fifteen different applications based on their role, and then ship them a new laptop based on where they live, plus, uh, make sure their benefits are up to date because they're a new employee," that is taking action. That is a complicated part. So even through MCP server or through any agent, they can give you the instruction. Now, the systems behind the scenes have to do that heavy lift. That's what we're solving for. Other than just giving you access, can I also now take action?

Carlos González de Villaumbrosia 24:15 And that leads me to the, the second bottleneck you identify, which is the scalability gap, right? Agents need a platform to perform. And one of the, the things that I've seen with, with product teams specifically is that there's always a group of people who are more tech-savvy than others, and they're able to set up their own agents. They, they, they find ways to increase their productivity by this multiplayer system that enables anyone ba-basically to use the agents to do something with the information without having to be the ones setting up the system. It seems very powerful, so I'm curious to know how, how you're seeing companies, uh, be able to get to that stage, multiplayer.

Amit Zavery 24:53 Yeah, yeah. See, as I said, if you had to build all this from scratch as individual companies, it will take a lot of effort and a lot of expertise. And, uh, just because you can wipe code an app doesn't solve the problem of scalability, doesn't solve the problem of, uh, really having agents, spawning new agents, doing multiple things, right? Because you're giving them access. They can go and run and operate for 24 by seven and keep on interacting and interfacing into those systems today. So the scalability gap which we see is that people are not rethinking how the usage will go away from human interaction to AI agent interaction. And if that is happening, the volume is changing, and the scalability of the platform is going to be very, very critical. So that's why what we're doing from the architecture perspective is making sure we can handle the volume, but also the security associated with that, right? So bringing the mindset of like, what is-- who's allowed to do what they're doing, how much volume you're allowed to do, what are the SLAs, and can you deliver against that? The customers have to solve that by looking at platform and their ability to scale to the volume you expect. And then you have to have a distributed architecture running in many environments and many zones, so that failovers are easier. Any kind of performance issue or bottleneck is taken care of. And architecturally, it keeps on evolving to be much faster and better performant, uh, uh, plat- uh, platform as well.

Carlos González de Villaumbrosia 26:22 Yeah. We're seeing early stages of this. I mean, Slack released their own bot, Claude integration with the Slack. And it seems like once the teams are able to work together across the same, the same system, then there's also this confusion around, okay, who can contribute to the ba-back to the system in terms of building new skills or maybe automating certain workflows? And then who is governing all of this? I think that's what you called out in the next, in the last two gaps, like wo-workflow gap and governance gap.

Amit Zavery 26:54 Yes.

Carlos González de Villaumbrosia 26:54 Like, how do you close that loop?

Amit Zavery 26:57 Yeah. See, the thing is governance is the biggest, uh, I would say, barrier to AI adoption. Uh, governance with security, if you want to think about that together.

Amit Zavery 27:08 But governance itself, because every large enterprise have to have a view about what is happening, where is it happening, can I-- do I n- have visibility, do I have control, and what risk am I taking, and how do you manage the risk? So the way we close the gap, we've been building governed and secured as well as risk management capability in our product and platform for a long time. Uh, we have a lot of key data and metrics being delivered to our users in terms of compliance, in terms of, uh, visibility, in terms of what is going on, uh, what applications were used, when they were used, how many licenses you have, what are, what is the pattern look like, who has, who has, uh, changes, what changes you want to make, the life cycle around it for human, non-human, physical AI, all of that stuff, right? So that is just a core feature in the platform we deliver. And as, as, as people are starting to build a lot of these applications, they should think about where do you run those things? What does the runtime look like? Is the runtime, has the right harness? Does it have the right kind of way of structuring information? And does it give you the right kind of reporting and mechanism to ensure that you're safe, you're doing this without risk? And if it's risk, what are the risks? Because your audit committee or your leadership team, everybody's going to ask you for that. How do you bring that kind of visibility into the picture? And that's how customers need to solve it, by having a platform which has that capability delivered as a core feature, not something on the side.

Carlos González de Villaumbrosia 28:42 You know, sometimes I've seen especially large enterprises using the word governance as an excuse, right? Because of course, there's so many things you have to regulate that by the time you're done with that, you'll never get it started with the actual-

Amit Zavery 28:53 Yeah, yeah

Carlos González de Villaumbrosia 28:53 ... implementation, right? So curious to know from your perspective, um, how can even large enterprises get started? You know, like make sure that people can still feel comfortable using the AI in certain ways, while of course there, there is a certain tower control, if you will, that is also going to create the right, the right guardrails without slowing down innovation.

Amit Zavery 29:14 Yeah. No, I think, uh, see, governance, I'm not a, n-never a big fan of putting policy for the sake of policy. And you don't want to have res- policies to restrict innovation and slow down adoption of new technologies. But you also want to have some safeguarding because some of these new technologies like AI and AI agents can really disrupt your business in a bad way by removing things or deleting things or doing things which are not allow exposing data. Uh, so every company needs to have some kind of policies, and they need to have some kind of, uh, governance structure to have that mindset and visibility. And I would not ex- uh, advise any customers to, say, put huge amount of barriers before you do anything, because that would be silly. But put some kind of structure that there are people and there are reports and there are tracking of things which are happening, so when things go wrong, you can stop it fast. Without visibility, you'd be too late to react. It'd be too late to fix, and your reputation as well as, uh, revenue, bo- and profit, all those thing can be compromised. So you do want to have, uh, and nothing wrong with some governance, but don't use it as an excuse to do less because I think we-- you're 100% right. I've seen company, a, a lot of, lot of leaders who say, "Oh, yeah, we have to put all these tools together. It'll take us three months or six months or nine months." Until then, nobody does any AI or nobody does cloud or nobody does this, and that is where the companies are going to die. But if you don't-- if, if you do governance with right kind of tooling while you accelerate your innovation, it's perfect, right? That's why we build those tools to remove that barrier. Say, "Hey, you have visibility. Let's go. Don't waste time. Uh, and we are not going to slow you down."

Carlos González de Villaumbrosia 31:07 I've seen two approaches to governance in, in, in enterprises. I've seen the, the centralized approach where there's like a, call it chief AI officer that is trying to kind of create the overarching policies and, uh, designing on the systems for all the functions. But I've also seen a more federated approach where each function has-

Amit Zavery 31:25 Yes

Carlos González de Villaumbrosia 31:25 ... their own system. And so that allows them obviously to go faster, but, like, the downside is it may be less control over the entire AI implementation. So I'm curious to know what you are seeing on your end.

Amit Zavery 31:36 Yeah. I've seen mix, to be honest. I think people are learning, so there's no, like, one size fits all. Every company has a different culture. Some companies do very well with central team because then they have one place to go, and some companies hate that centralization. That means I have somebody else telling me what to do. I don't like it. So it depends on the company culture. So I don't think so there's anything wrong with either model as long as you're clear what your model is. You cannot have everything in different, different ways. You cannot have a chief AI officer with a central team, everybody federated, nobody listening to each other, and it'll be a chaos. So I think it's okay to have structure, and I think it's neither of the structures are wrong if you have the right people with the right mindset and willingness to change. If you find something not working, you have to be able to change it. I've-- What we have done at ServiceNow, every com- every group has to be able to do AI. So we don't have this idea that it has to be all centrally monitored and managed. But we provide s- we have created in every department a COE, which is loosely kind of defining what a department like finance department or, uh, our HR group should be doing. They have the COEs, which are AI COEs, very experts, very senior leader running that, and they work with other COEs in a federated model. While if there's some issues, because we are a product and engineering company, if there anything we need to do centrally, we can bring that expertise as needed to allow our IT team is kind of acting like a central clearing house. "Hey, this project, let's do it. Let's kind of implement it. Let's get it going." But the ideas and ability to do it all happens in every department. So there's a lot of good ways to move fast, but there's also ways to kind of get help if you need to and not go down the wrong path by not understanding it, right? So I think people have to just put some structure. I think it's a good thing because these things are moving fast. It's changing fast. There's a lot of risk associated with that, so putting a right kind of structure is important, I think. But don't over-- go overboard trying to find the best structure.

Carlos González de Villaumbrosia 33:45 And ultimately, all of this comes down to ROI, right?

Amit Zavery 33:48 Yep. Yeah.

Carlos González de Villaumbrosia 33:48 And then so you mentioned even in your report that these pacesetters are able to get ROI over 100%. I think you said 160%, something like that. So I'm curious to know, okay, regardless of the approach you take to increase AI adoption and ultimately ROI, what are some of the good metrics that help you prove your point and, and show the team that we are on the right path there?

Amit Zavery 34:11 See, I think, again, it's similar to the structure. ROIs are in the eyes of the company in terms of what they value. So if, if, if, if it's all about, uh, bottom line or improving efficiency or speed, uh, you'll use those metrics. We see a lot of metrics we provide to our customers, both in terms of, uh, uh, time to value, time to, uh, resolve an issue, uh, time to, uh, uh, fix something, uh, metrics like the number of hours saved in automation, uh, amount of processes cleaned up so that you used to ta- they do thousands of, uh, different variations. How are you simplifying? So we provide a lot of those metrics, uh, for our customers, and they pick which helps them see the ROI on it. Of course, the amount of money they're spending and the number of tokens may be using, all those things are all good metrics, good input to the outcome. The-- It-- No, no- nobody should just worry about one minor item. You should be looking at holistically. And what we're talking about metrics are the more corporate level metrics, and that's the one I would suggest for anybody to calculate ROI is that are they getting savings? Are they getting automation? Are they get-- becoming more efficient? Are they resolving things faster? And that, and eventually all shows up in the top line and bottom line. If your revenue is growing, and are you, are you becoming more efficient? And that are the ones which I think people I talk to worry about more. If they don't see that, then they don't believe it. They say, "Well, you're doing this thing for the sake of doing it, and I'm not getting any, uh, top line and bottom line improvements."

Carlos González de Villaumbrosia 35:48 I want to talk about go-to-market because I think there's a lot of companies who are already on the right path, right? They bought the tools. They're encouraging the teams to use AI. They are seeing some productivity gains, but still they, they might need some external help. And I, and I notice in, in your model, you have forward deployed engineers. But you also have, it seems like a marketplace called Autonomous Workforce.

Amit Zavery 36:10 Yeah.

Carlos González de Villaumbrosia 36:10 So I'm curious to know about those different approaches you are taking to help more organizations adopt AI.

Amit Zavery 36:16 Yeah. I think, see, the, uh, when we started doing agentic, uh, last year, one of the biggest barriers for agentic-- Well, two barriers. One was worry about governance and security. It was the number one issue for everybody. They didn't know what will happen, what datas will be accessed, what system will be touched, can, can it be prevented from doing something wrong, all that kind of stuff, right? So visibility, control became number one, and that's why we launched a product called AI Control Tower, to give you full visibility about all your AI systems, the cost, the life cycle, the versioning, uh, access, security, all that stuff in one place, the observability and everything else. The second, uh, thing for, uh, a lot of, lot of the customers we're talking to is that, where do I start? And, and that's where the go-to-market team started getting more engaged to show them all the different things you can do with our agentic processes. Take your incident management process or triage process or, uh, case resolution process, uh, security, uh, the vulnerability process. All these processes and business, business flows and workflows, uh, you needed to figure out where to start and where you're gonna get the best ROI. This is where we put in investment in our solution consultant as well as FDE to show customers real value of our products by implementing one of those pro-- workflows and agentifying what they might have used before into an AI-based, uh, outcome, and out of the box, easy to do, and you don't have to rewrite everything every day. So that's really where we saw a lot of good traction because we took the k- uh, issue out of security and compliance out of the table by giving AI Control Tower, and then, uh, making a lot of automation happen through, uh, playbooks with agentic, full-blown AI capabilities available out of the box, the 100 plus like that. So those two things got us huge adoption because customers are liking it. They know how to do it. We were able to help them with the first few use cases, and they wanted more and more after that. And then we realized, see, building AI agents is not the end goal of every company. Why do they care? They care about a solution. They care about doing the job better and faster. If an AI agent's doing it or human agents are doing it, for the end user, it doesn't matter really, typically. So that's why we introduced this idea of autonomous workers, where we can take a full-blown ti-- uh, uh, work, be L1 support engineer, SecOp analyst, HR business partner, their full job, and provide that as an autonomous worker doing it end-to-end and taking actions for you so that it offloads from human, reduces the time to process, so we can go from, like in general, it should be two days to twenty minutes, as well as you can now close a lot of the issues autonomously without human interaction. So it saves you money and time, but this is running twenty-four by seven with multi-language. You don't have to build huge systems out there, people doing all this work. Now I can operate that with autonomous. So that's the w- that's the marketplace we're creating with this autonomous worker. We have 20 of these who are now taking care of all these end-to-end tasks in using AI underneath, but customers don't have to worry about it. So we're moving away from this idea that you don't have to do every AI agent every day. We are giving you the full work capability.

Carlos González de Villaumbrosia 39:38 So, so to be clear, these are humans. Obviously, they might be using agents, but the, the, the companies are, are hiring these humans, and this is a marketplace. This is different than your own in-house forward deployed engineers, right?

Amit Zavery 39:51 Yeah. Well, so there are two parts. The forward deployed engineers are helping customers with generic AI-related stuff. Like they go help them from week or two weeks and take a product. It could be, uh, autonomous workers. It could be AI Control Tower. It could be the agentic workflows. Depending on the customer need, we get them started. Autonomous, uh, AI specialists are full digital equivalent of human workers. So L1 support engineer can-- a human agent can be replaced with this autonomous L1 support engineer, uh, uh, AI specialist, and do the same task with the same skills but do it faster, do it in a, uh, in, in a much more efficient way. And that FDEs can also help you with that. Take this autonomous worker and make it work in your environment. So the FDEs are really making AI work for our customers, and the technologies like AI Control Tower, autonomous worker, agentic platform are the tools or the full skills and the solutions for customers to really be successful.

Carlos González de Villaumbrosia 40:53 It makes sense because in this agentic world, a lot of SaaS companies are repositioning themselves as the agentic layer.

Amit Zavery 40:59 Yep.

Carlos González de Villaumbrosia 40:59 Everybody wants to be the layer on top of all your tools, workflows, data, LLMs, you name it, right? Orchestrate AI across the company. And that's a bold bet because, uh, if you, if you win, you win big, but everybody wants to do it, right? So I'm curious to know from your perspective, what makes you believe, what's your angle to eventually become that, that, that agentic layer on top of everything else?

Amit Zavery 41:23 Yes. So ServiceNow has always been the orchestrat- orchestrator for business processes end to end, right? So inside an enterprise, when we connect various system together, ServiceNow is used as an agent-- as a, I would say, orchestration layer before even agents came along, right? So we've been in that business for a long, long time, and we have experience connecting and automating business processes for years, twenty-plus years. We definitely have ability to do orchestration, but I don't think, and I think the industry thinks there's going to be one orchestrator, which is a fallacy. There are going to be multiple businesses, business processes orchestrated by different, different providers. Uh, and, uh, in some cases, different orchestrators have to work with each other. Some cases, they might be siloed because they're only doing things in a particular department or particular application. Uh, so my guess and my expectation is over time is that every company will do some orchestration. We might do a lot more than everybody, but we might not do all of it. It has to, like in some places, if I'm using a vertical stack, say I'm doing an industry application in, say, health sciences, and that application will do the orchestration for those agents in that environment. And then we would come along, and we'll do orchestration of our agents or third-party agents when we're building a business process which touches that application and connect them together, right? Uh, because that application is very verticalized in their own world, they don't have any visibility across outside. We would connect and orchestrate that agent system with a third party, other agent, where we might need a data flow from one place to another. So if you have source to pay or order to cash, all of these things are multiple systems out there, and we've been doing orchestration for a lot of those core running processes, and we will do the same thing in the AI world, while some of the other parts of those processes will be run by somebody else. So I don't think I'm as expecting or I'm saying to any customer that we are going to be the Uber of ev- Uber orchestrator of everything, and there's nobody else who can ev- ever be as well because I think people are thinking that they will own the whole enterprise. That is never going to happen. Enterprise is very fragmented. They have a lot of legacy, a lot of new systems, a lot of bespoke applications, a lot of vertical stack, a lot of, uh, new AI systems out there. You will be part of it, and you have to always think about that, how do I make it work across, instead of trying to say, "I am in this small world, and I-- everybody has to be with me." That will be a, a failure mode for every comp... If every-- any company who wakes up and says, "I am the end-to-end orchestrator," I think they're dead.

Carlos González de Villaumbrosia 43:57 I, I've seen that movie before, and I agree. Uh, when everybody w- was trying to replace Excel, and they would put themselves in the center of the picture and say, "And we integrate with everybody else." And then the reality is that Excel is still alive. There's a bunch of SaaS tools that are still necessary that integrate with Excel, and there are multiple winners.

Amit Zavery 44:12 Yeah. So I think that's one thing that we cannot-- You have to realize, and if you are in enterprise space, uh, uh, enterprise software, you have to know how enterprises work. These businesses are never going to completely do everything end-to-end with one thing because it's not possible. You have to have multiple things working together, and that's why we always have believed in building an open ecosystem. We've never been a provider says everything is ServiceNow. Our, our job is to make it work across. If you come along and say, "I will o-- I will be the top of the spear, and I will be the only one. Everybody else work with me," your business will be dead.

Carlos González de Villaumbrosia 44:50 Yet I, I notice you are also expanding your own applications, right? So you recently acquired a cybersecurity business for over seven billion dollars. I, I notice you also launched a new s-- uh, a-autonomous CRM. So I'm going to try to touch on, on those, starting with cyber, right? Like, what is your thesis or rationale behind this investment?

Amit Zavery 45:11 Since a lot of people don't know, our security business is billion-dollar plus before I even did the acquisition, right? So our security b- uh, business for doing any kind of CISO-related post-breach activity. A CISO is our second-largest buyer for ServiceNow after CIOs, and the reason they are is because they use us with any s-security tool out there, Spalo, CrowdStrike, uh, Google, Wiz, whatever it is. Back end, in terms of integrating with a workflow for doing any kind of TRG or incident management, it's ServiceNow. So we already have been working with CI, uh, CISOs for a long time. As, as we saw AI, we saw an opportunity accelerate our security roadmap i, uh, uh, in the areas of non-human identities and identity governance because we were governing humans' identity inside ServiceNow today. We are the ones, we onboard, offboard employees typically for every company, what access they have, what a- what systems they're running, what VMs they're using. We used to know all that stuff in our system today in CMDB. So expanding that to non-human identity was very natural for us. So that was the way the Veza is a small company, but a very innovative a-access graph company for identity governance, and it's really making a difference, especially when the volume of, uh, AI agents are growing. You need some kind of governance around it. So that's the product we provide now, part of AI Control Tower and our security portfolio. Second thing which is happening is, of course, we work very, very, very well with IT. We also were starting to work very well with OT, the operational technology, the shop floor, manufacturing, devices. They also had issues and incidents, and they use ServiceNow to resolve them. So IT and OT is starting to become one platform for management. So we were providing them that technology already, and the company Armis, which is the number one provider in OT cybersecurity and exposure management, we bought them so that we can close the gap and speed up our roadmap and create a bigger TAM for ServiceNow on this exposure management. If some exposure you have in your devices, how do you manage it? What are the-- How are you gonna resolve it? How are you gonna take it offline? Because anything in your shop floor, in your factory, if it goes wrong, can bring down the factory. You will be out of business for a few weeks or a few months, which is not great for any company. That's where Armis fits in, is to really provide this IT, OT, cybersecurity exposure management or vulnerability management, uh, using AI on top of what we were doing already. And now that is great because now customers are like, "You know what? I want to be able to have my IoT devices, OT devices, medical devices, physical AI all monitored and managed, uh, through, uh, through ServiceNow." So that's where it fits in for us. So we've been very thoughtful about where we get in. We're not doing everything. Cybersecurity is huge space. But for us, this area where we were and expanding it into new TAM, which is very natural for us, makes sense.

Carlos González de Villaumbrosia 48:16 Yeah. No, now that you, you, you explain it, I think it makes sense. It's a, it's a natural adjacency to IT, especially for OT departments. Now, the other expansion that I noticed you made is this autonomous CRM.

Amit Zavery 48:30 Yes.

Carlos González de Villaumbrosia 48:30 That to me sounds less obvious, at least from the outside, right? You're getting into the sales teams. You are competing directly with giants like Salesforce. So what is the rationale behind it?

Amit Zavery 48:40 Yeah, I think C- CRM is multiple businesses definition-wise. I think people confuse, uh, CRM as only sales. There are three major parts, or four mo- major parts of CRM, uh, sales, service, marketing, and commerce. The parts which we think we have natural opportunity for us is be- is definitely for sure s- uh, customer service. Customer service is similar to any incident management we do for employees. When employees have issue with IT, they file a ticket or they call someone to get help. Now they go through chatbot or whatever it is. They're asking for help, and we are the ones, ServiceNow was the one which is resolving that issue and helping them fix it. So taking that whole case management and incident management, when a customer now, end customer calls you saying, "Uh, my order is missing," or, uh, or, "I want to change something in my account," or, uh, "This is broken. How can you fix it?" It's same thing as case management we did for employees. Sure, it goes through telephony network and all kind of stuff, which we integrate into. So for us to move from, uh, what we were doing in internal employee to cust- end customer customer, especially for customer support, was very natural. It's the same understanding. We know how to make it work. We understand how to orchestrate between different users and com- uh, people who need to fix the issue and resolve it for the end customer. And our sec- uh, our, our, our CRM business in customer service has become huge now because we've been able to build very solid product on the same platform, very modern, very AI-driven, wh- as you said, autonomous, and also now omni-channel, understanding voice, chat, documents, text, uh, web, all that stuff through one int- through one product underneath the covers, so you don't have fragmentation. So we can differentiate ourself against all the incumbents in that space, and the business has crossed over 2 billion because it's been on a very fast trajectory. So we are able to do very well in that area because we have a great product. And related to that, anything which requires complex orchestration, uh, like CPQ, com- uh, configure price quote, it requires orchestration between a configuration from a sales team to a pricing team, a legal team, and getting a quote out based on product. So CPQ is a very natural interaction for CRM for us. So the things we are doing are areas where we have a lot of expertise, a lot of capability, and ability to really differentiate and help our customers move fast. So we are open to competing. I mean, there's no harm. I mean, if you build a great product, uh, we think we can win, we keep on winning. Uh, and, uh, we don't do things like... I mean, we allow, we of course provide products in Salesforce automation with sales order management, but the things are just about finding information, like give me the forecast. It's not complex, and that is not something which we really are doing a lot of work in. The part which we do is order management, uh, sales efficiency, uh, doing customer support, field service management, CPQ, which are complex orchestration and giving you a outcome. Like when you do CPQ, you need to get a quote out. When you do order management, you need to get order finished. Uh, when you are doing customer support, you want to resolve an issue. Those are things, outcome-driven things we are very, very excited about participating because we have the platform and the ability already to do it. But we're not going into random space like marketing or things like that. We partner in those areas.

Carlos González de Villaumbrosia 52:17 So it's more like the, the focus on the, the, the use case around CRM for customer service that has a stronger connection to what you do already with IT and OT. Got it.

Amit Zavery 52:26 Yes, yes.

Carlos González de Villaumbrosia 52:26 I mean, and, and la- last thing on this, I mean, one of the modes for, for CRMs has been high switching cost.

Amit Zavery 52:33 Yes.

Carlos González de Villaumbrosia 52:33 Still with AI, I think the cost of switching is still pretty high, right? So i- in your case, even though you're already a, a customer, you might be o- owning IT, HR, or other, other functions. Like, what is the motion there to eventually replace and, and gain more market share on the CRM side?

Amit Zavery 52:51 So as I said, we are doing very well in the CRM side, so that means things we are doing is working. We help customers migrate. We have tooling. We guarantee them the time. We have a lot of partners we work with as well to help with the migration if they need to. Sometimes there's integration work, not just migration, because people might say, "I want to modernize some parts of it," and then modernize everything over time. Uh, we also provide them support from licensing perspective, so they don't have to worry about, uh, uh, things running two different systems together for too long. We can migrate it faster. Uh, we have, uh, done a lot of good work to be able to migrate from all the legacy CRM systems, customer support systems to us. Uh, today we have a lot of successful customers now, uh, hundred- uh, thousands of customers who are on, uh, ServiceNow's, uh, uh, CSM, uh, and, uh, we working aggressively to keep on getting them faster and faster.

Carlos González de Villaumbrosia 53:44 Well, Amit, thank you for this, uh, full rundown on AI transformation-

Amit Zavery 53:49 Yeah

Carlos González de Villaumbrosia 53:49 ... and, and product expansion. Uh, it's been a pleasure to learn and spend this time with you.

Amit Zavery 53:54 Carlos, we are very busy, but thank you for taking, uh, taking time with me as well about, uh, all the things we are doing so that your, uh, listeners can l- learn about ServiceNow and how we can help them as well, so. [outro jingle]